EAAT2 and the molecular signature of amyotrophic lateral sclerosis

Abstract

Amyotrophic lateral sclerosis (ALS) is a rapid and fatal neurodegenerative disease, primarily affecting upper and lower motor neurons. It is an extremely heterogeneous disease in both cause and symptom development and its mechanisms of pathogenesis remain largely unknown. Excitotoxicity, a process caused by excessive glutamate signaling, is believed to play a substantial role, however. Excessive glutamate release, changes in post-synaptic glutamate receptors, and reduction of functional astrocytic glutamate transporters contribute to excitotoxicity in ALS. Here, we explore the roles of each, with a particular emphasis on glutamate transporters and attempts to increase them as therapy for ALS. Screening strategies have been employed to find compounds that increase functional excitatory amino acid transporter EAAT2 (GLT1), which is responsible for the vast majority of glutamate clearance. One such compound, ceftriaxone, was recently tested in clinical trials but unfortunately did not modify disease, though its effect on EAAT2 expression in patients was not measured.

Amyotrophic lateral sclerosis (ALS) is a rapidly progressive, fatal neurodegenerative disease that primarily affects motor pathways. Upper and lower motor neurons of the corticospinal tract selectively degenerate leading to muscle spasticity, weakness, muscle wasting, and death. With an incidence of around 1.75 per 100,000 people per year, ALS typically leads to death from respiratory failure in only 3–5 years after symptom onset, though about 10% of patients live substantially longer (Marin et al. 2016; Turner et al. 2013). Diagnosis is made at an average age of 55 and is somewhat more common in men. ALS is a highly variable disease: Patients can present with limb (70%), bulbar (25%), or trunk/respiratory (5%) onset, with a combination of upper and motor neuron symptoms such as spasticity, muscle fasciculations, cramps, weakness, and muscle atrophy (Ross and Tabrizi 2011; Turner et al. 2013). About 10–20% of cases are familial, caused by mutations in over a dozen genes (Robberecht and Philips 2013). Around 20% of familial cases are caused by mutations in the SOD1 gene, one of the first discovered causes. Mutations in the RNA binding proteins TDP-43 and FUS account for 1–5% of familial cases each. A six-hexanucleotide (GGGGCC) repeat sequence in C9orf72 was found in 2011 and causes ~40% of familial ALS, as well as a portion of sporadic cases. The majority of inciting factors for sporadic ALS remain unknown.

SOD1 mutations have been used to generate many transgenic animal models of the disease, including the first human SOD1-G93A transgenic mouse model, which have been used extensively over the last two decades (Gurney et al. 1994; Rosen et al. 1993). The SOD1-G93A mouse model of ALS recapitulates many of the symptoms of ALS, as mice first demonstrate a hindlimb tremor, followed by loss of the hindlimb splaying reflex, and eventually paralysis and death. Generation of chimeric mice comprised of mixtures of normal and SOD1 mutant expressing cells demonstrated that glia cells contribute to the death of motor neurons in a process known as non-cell autonomous toxicity (Robberecht and Philips 2013). Astrocytes expressing mutant SOD1 and astrocytes derived from sporadic ALS patients release factors toxic to motor neurons (Haidet-Phillips et al. 2011; Nagai et al. 2007). Unfortunately, while the SOD1 models have dramatically increased our understanding of ALS, they are based off a small fraction of an extremely heterogeneous disease.

Perhaps because of this heterogeneity, the exact pathways of motoneuron toxicity have yet to be understood. Excitotoxicity, however, is believed to play a role. Critical to central nervous system function, the neurotransmitter glutamate communicates the majority of excitatory signals across neuronal synapses. Post-synaptic neurons process its message through a variety of ionotropic and metabotropic glutamate receptors, while glutamate transporters located primarily on astrocytes surrounding the synapse remove it, ending the signal. When this process breaks down, glutamate overstimulates the post-synaptic neurons. This in turn leads to excessive Ca²⁺ influx through NMDA receptors, some AMPA and kainate receptors, and voltage-gated Ca²⁺ channels (Van Den Bosch et al. 2006). The Ca²⁺ burden can cause mitochondrial damage, enzyme activation, reactive oxygen species generation and other damaging processes. The neuronal damage caused by excessive stimulation, known as excitotoxicity, has been implicated in a variety of acute and chronic conditions, including ALS, epilepsy, cerebral ischemia, schizophrenia, mood disorders, and anxiety (Lauriat and McInnes 2007).

Impaired glutamate transport leads to increased synaptic glutamate, which can lead to excitotoxicity. Toxicity is primarily caused by excessive stimulation of receptors, leading to Ca²⁺ influx and dysregulation. This leads to mitochondrial uptake and dysfunction as well as other organelle dysregulation. Additionally, excessive glutamate signaling leads to TF & IEG activation, protease and calpain activation, cytoskeletal alterations, and ROS production. For a detailed review of its mechanisms, the readers are referred to an earlier review by Wang and Qin (Wang and Qin 2010).

Ample evidence suggests that excitotoxicity, especially of motor neurons, can be induced by impairment of the glutamate transport system. Loss of EAAT1 (GLAST) or EAAT2 (GLT1) by antisense oligonucleotide knock-down in organotypic spinal cord cultures cause a progressive decline in motor neuron viability as measured by ChAT activity (Rothstein 1996). Knock-down of the same astroglial transporters in vivo led to a rapid, progressive motor neuron syndrome (Rothstein 1996). Mice lacking EAAT2 were more susceptible to edema following cold-induced brain injury, with 68% greater edema (Tanaka et al. 1997). Riluzole, the only FDA-approved medication for the treatment of ALS primarily functions by inhibiting persistent Na⁺ currents, repetitive neuronal firing, and neurotransmitter release (Bellingham 2011). Those functions would presumably act to attenuate excitotoxicity by decreasing the excitation of the post-synaptic neurons.

While glutamate is normally cleared from the synapse by excitatory amino acid transporters, the highly predominant transporter, EAAT2 is substantially decreased in ALS. (Bruijn et al. 1997). Motor neurons also appear to be particularly vulnerable to excitotoxicity (Kuner et al. 2005). Furthermore, riluzole, the only FDA-approved drug for treating ALS has been shown to inhibit excitotoxic pathways (Bellingham 2011). EAAT2 may play a role in other neurotoxic pathways in ALS, as the formation and accumulation of a sumoylated fragment has been shown to induce the secretion of neurotoxic factors by astrocytes (Foran et al. 2011). In the last decade, substantial research has been directed towards developing pharmaceuticals that block excitotoxicity and/or increase EAAT2. Specifically, a variety of screening platforms have been employed to detect compounds that are capable of increasing EAAT2 with the goal of improving glutamate clearance and preventing excessive signaling. A number of compounds, including the blood brain barrier (BBB)-permeable ceftriaxone, have shown promise in cell-based assays and in animal models of ALS (Colton et al. 2010; Li et al. 2011; Rothstein et al. 2005). The recent failure of ceftriaxone in clinical trials was a significant set-back but as EAAT2 changes were not actually measured, EAAT2 remains a viable clinical target for the treatment of ALS (Cudkowicz et al. 2014).

Glutamate dysregulation in ALS

Dysregulation of glutamate levels in ALS has been well explored, though studies have offered conflicting reports of plasma and CNS glutamate levels, likely due to differences in methods and subgroup composition. It has also been reported that glutamate measurements are dependent on sampling variables, particularly storage temperature, perhaps accounting for discrepancies (Wuolikainen et al. 2011). Plasma glutamate levels were substantially elevated and oral glutamate loading resulted in greater levels in early stage ALS patients when compared to healthy and diseased controls (Plaitakis and Caroscio 1987). Even more compelling, glutamate was elevated in the cerebrospinal fluid of ALS patients by 100–200% (Rothstein et al. 1990). Conversely, glutamate levels were reduced in the CSF of patients as analyzed with GC/TOFMS (Wuolikainen et al. 2011). Glutamate levels were also found to be decreased in the frontal cortex, cerebellar cortex, lumbar spinal cord, and cervical spinal cord of patients who died of ALS, though the decrease may have been a result of neuronal loss or it could demonstrate the dysregulation of glutamate (Plaitakis et al. 1988). The distinction could also reflect a difference in patient composition: Camu et. al. found glutamate to be elevated in the plasma and CSF of ALS patients with spinal onset but decreased in the plasma of ALS patients with bulbar onset (Camu et al. 1993). Similarly, plasma glutamate levels were elevated in ALS patients, but only in spinal onset patients (Andreadou et al. 2008). In a large cohort, 41% of ALS patients demonstrated high glutamate levels in the spinal cord, which correlated with spinal onset, more impaired limb function, and faster muscle deterioration (Spreux-Varoquaux et al. 2002).

Glutamate release from spinal cord synaptosomes was significantly higher in SOD1-G93A mice, both at basal levels and with KCl- or ionomycin-evoked release (Bonifacino et al. 2016; Milanese et al. 2011). While the evoked response was more pronounced at later disease stages (over 120 days of age), it was observable even at the early presymptomatic stage (at 30–40 days of age). Basal efflux was even more pronounced at the earlier ages, probably due to a larger readily releasable pool of vesicles. Activation of mGlu1/5 receptors (Group I receptors) elicited abnormal glutamate release in synaptosomes of SOD1-G93A mice via calcium release through IP3-sensitive channels (Giribaldi et al. 2013).

Glutamate receptors changes in ALS & motor neuron vulnerability

As the primary excitatory neurotransmitter of the central nervous system, glutamate signals to post-synaptic neurons through both receptor-gated ionotropic and G-protein coupled metabotropic receptors (mGluRs) (Nicoll et al. 1990). The ionotroptic receptors, including NMDA, AMPA, and kainate receptors, are all permeable to Na⁺ and K⁺, but only NMDARs are typically Ca²⁺ permeable as well (Dingledine et al. 1999). AMPA receptors are composed of GluA subunits which confer different properties. Usually included, the GluA2 subunit is typically modified at the mRNA level, leading to a glutamine to arginine mutation and switching the final receptor from calcium-permeable to impermeable (Kawahara et al. 2004). AMPA and kainate receptors without the GluR2 subunit or unedited in the Q/R site of the M2 region are Ca²⁺ permeable as well, though much less prevalent. NMDARs are also susceptible to external Mg²⁺ blockage relieved by depolarization, which can be caused by activation of AMPA and kainite receptors.

In ALS, motor neurons appear to be selectively vulnerable to excitotoxicity, particularly via AMPA and kainate receptors. While cerebrospinal fluid (CSF) of ALS patients is toxic to motor neuron cultures, that toxicity can be prevented with AMPA and kainate receptor blockers such as CNQX, but not by NMDA antagonists (Couratier et al. 1993; Sen et al. 2005). AMPA and kainate antagonists such as NBQX have also been shown to decrease toxicity in cell models and improve disease progression in the SOD1-G93A mouse model of ALS (Anneser et al. 2006; Tortarolo et al. 2006; Van Damme et al. 2003; Yin et al. 2007). In support of the AMPA/kainate receptor mediated pathway, exogenous AMPA and kainate have been shown to cause motor neuron toxicity (Carriedo et al. 1996; Corona and Tapia 2004). Infusion of kainic acid to the lumbar spinal cord leads to a dose- and time-dependent loss of motor neurons and to alterations and loss of distal neuromuscular junctions (Blizzard et al. 2015). Furthermore, iPS motor neurons derived from C9orf72 patients were 100-fold more sensitive to glutamate excitotoxicity (Donnelly et al. 2013). Toxicity was blocked by GluA and calcium channel inhibitors. This might be due to loss of ADARB2 via sequestration in RNA foci. A compartmentalized model of excitotoxic exposure was used to show that somatodendritic but not axonal exposure to kainic acid was highly toxic to mouse lower motor neurons in vitro (Blizzard et al. 2015).

The selective vulnerability of motor neurons is likely due to changes in AMPA receptor subunit composition seen in ALS. While in normal cells, NMDA receptors are the primary calcium-permeable receptors, AMPA receptors can also be so under certain circumstances. AMPA receptors are composed of a tetramer of subunits which usually includes GluA2. The GluA2 mRNA is post-transcriptionally edited by adenosine deaminase acting on RNA 2 (ADAR2) in the second transmembrane region, causing an R to Q mutation in the coded protein at codon 607, a mutation which changes the subunit from being calcium permeable to impermeable (Kuner et al. 2005). Normally, the editing efficiency is near complete, so that virtually all AMPA receptors containing GluA2 are calcium impermeable. Alternatively, if the receptor includes GluA3 instead of GluA2 or is unedited, it can be calcium permeable. It has been found that in the motor neurons of ALS patients, GluA2 is incompletely edited at the Q/R site (almost half of patients, ranging from 0–100% editing) in motor neurons of ALS patients, leading to increased calcium permeability (Damme et al. 2002; Kawahara et al. 2004; Takuma et al. 1999). The GluA2 editing enzyme ADAR2 was absent in more than half of sporadic patients and none of controls, coinciding with phosphorylated TDP-43 inclusions (frequently observed in ALS) in neuronal nuclei (Aizawa et al.2010). A later study confirmed ADAR2 was downregulated in all sporadic patients studied and all had motor neurons expressing unedited GluA2 versus no motor neurons with substantially unedited GluA2 in controls (66.0 ±22.7% versus 99.4 ± 0.7% of GluA2 edited) (Hideyama et al. 2012). In an unfortunate cycle, excitotoxic glutamate or NMDA exposure caused neurons to cleave ADAR2 in a calcium- and calpain-dependent pathway, preventing GluA2 editing (Mahajan et al. 2011). It is possible that AMPA could have a similar effect once GluA2 becomes unedited, as it would also become calcium permeable. A decrease in GluA2 expression relative to GluA3 has also been seen in SOD1-G93A mice (Tortarolo et al. 2006).

Other glutamate receptors also play a role in ALS. Metabotropic glutamate receptors are categorized into three groups: Group I includes mGlu1 and mGlu5, which are G_q receptors whose activation leads to IP₃ formation; Group II includes mGlu2 and mGlu3 which are G_i/o receptors and largely function by inhibition of adenylyl cyclase; and Group III includes mGlu4, 5, 7, and 8, which are also G_i/o receptors. It was found that modulation of mGlu group 1 receptors was protective of chicken embryonic spinal cord cultures from ALS-CSF-induced toxicity (Anneser et al. 2006). Supporting the role of mGlu in excitotoxicity, SOD1-G93A mice with only one copy of the gene coding for mGlu1 exhibited delayed onset and slower progression, associated with a decrease in abnormal glutamate release (Milanese et al. 2011). Stimulation of mGlu3 but not mGlu2 caused GDNF secretion which protected spinal motor neurons in mixed cultures from excitotoxic death (Battaglia et al. 2015). A group II agonist also increased EAAT2 expression in the spinal cord and rescued spinal cord motor neurons in the SOD1-G93A mouse model but had no effect on the lifespan of the mouse.

Glutamate transporters & dysregulation in ALS

Astrocytes play a crucial role in neuronal signaling. In one prime example of the integrated neuron-glia relationship, while a presynaptic neuron signals to a postsynaptic neuron with the excitatory amino acid, glutamate, astrocytes are primarily responsible for clearing glutamate from the synapse to terminate the signal. Synapses depend on the neurotransmitter signal, its activation of receptors (both pre- and post-synaptically), and the termination of that signal. Astrocytes clear glutamate from the synapse with excitatory amino acid/glutamate transporters (EAATs) to end the synaptic transmission signal. Five excitatory amino acid transporters can be found in the human CNS, each with distinct regional specificities and physiologic properties (Arriza et al. 1994; Fairman et al. 1995). Isolated and cloned first, rodent homologues demonstrated sodium-dependent and chloride-independent uptake of glutamate and aspartate. EAAT1, or GLAST in rodents, predominates in astroglia of the cerebellum but can be found elsewhere in the cerebrum as well as other tissue including the heart, lung, and muscle (Arriza et al. 1994; Rothstein et al. 1992; Storck et al. 1992). In contrast, EAAT2, or GLT1 in rodents, is found exclusively in astroglia of the CNS and minimally in the cerebellum and other tissues (Arriza et al. 1994; Danbolt et al. 1990; Pines et al. 1992). EAAT3, or EAAC1 in rodents, is distributed throughout neurons of the cortex and non-CNS tissues including the lung and kidney (Arriza et al. 1997; Kanai and Hediger 1992; Rothstein et al. 1994). EAAT4, cloned in 1995, is found predominantly in cerebellar Purkinje cells, while EAAT5 is found almost exclusively in the retina (Arriza et al. 1997; Fairman et al. 1995). In antisense oligonucleotide knock-down experiments, EAAT1 (GLAST), EAAT2 (GLT1), and EAAT3 (EAAC1) were found to be responsible for approximately 30%, 50%, and 20% of glutamate transport in the striatum, respectively (Rothstein et al. 1996). Concordantly, impairment of glutamate transport led to 13- and 32-times the normal glutamate levels with knock-down of only the astroglial transporters, EAAT1 (GLAST) and EAAT2 (GLT1).

EAAT2 (GLT1 in rodents) is responsible approximately 90% of glutamate clearance from the synapse (Lauriat and McInnes 2007). Because human EAAT2 and rodent GLT1 have 96% homologous sequences, EAAT2 is often used to denote both human and rodent protein, as is the case for this chapter. A similarly high degree of homology is shared by the rodent glutamate transporters, GLAST and EAAC1, with their human counterparts, EAAT1 and EAAT3 (Kirschner et al. 1994). Synaptic glutamate rapidly binds to it (with a rate constant of 10⁻⁷/M/s) and then is more slowly transported into the astrocytic space (at about 30 molecules/s) (Takahashi et al. 2015). Three Na⁺ ions and one H⁺ ion are cotransported and one K⁺ ion is countertransported with each glutamate molecule, with transport coupled to the sodium gradient. The EAATs also function as selective anion channels, with chloride permeation gated by the lateral movement of the glutamate transport domain and pore hydration (Machtens et al. 2015). EAAT2 has eight transmembrane domains and a long 3’ UTR for regulation. Multiple isoforms of EAAT2 have been discovered, primarily classified by their various C-termini as the originally isolated astroglial EAAT2a, the less prevalent astroglial EAAT2b, and the retinally expressed EAAT2c (Chen et al. 2002; Holmseth et al. 2009; Lauriat and McInnes 2007). EAAT2b and EAAT2c have truncated C termini with instead alternative 11 amino acid sequences. All three variants have some differences in the 3’-UTR as well, suggesting differential regulation (Lauriat and McInnes 2007). All increase towards adulthood, though EAAT2a more than others (Holmseth et al., 2009). Complete knock-out of EAAT2 in a mouse model caused premature death (50% survival after 6 weeks) following spontaneous epileptic seizures (Tanaka et al. 1997). Cortical synaptosomes had only 5.8% of the glutamate uptake capacity as controls, suggesting that EAAT2 is responsible for over 90% of glutamate clearance in the cortex. Peak concentrations of synaptically released glutamate in tissue slices were increased and remained elevated in the knock-out mice, indicating impaired glutamate clearance following neurotransmission.

Essential to the functioning nervous system, the excitatory amino acid glutamate and its transporter EAAT2 are associated with multiple disorders, including amyotrophic lateral sclerosis (ALS), stroke, Parkinson’s Disease, and epilepsy (Takahashi et al. 2015). Understanding the regulation of EAAT2 expression could yield clues about its dysregulation in disease states and provide targets for altering it in therapy. Similar regulation of EAAT2 expression is seen in rodents and human astrocytes: Modulators include EGF, cAMP, PACAP, TGF-β, TNF-α, ceftriaxone, and estrogen compounds, often functioning through NF-κB (Takahashi et al. 2015). The transcription factor Pax6 also interacts with a distal enhancer element to increase EAAT2 transcription in astrocytes (Ghosh et al. 2016). Neuronal regulation also plays a role via presynaptic terminals (Yang et al. 2009). Neuronal soluble factors have also been found to induce EAAT2 expression in astrocytes via RTK signaling (Gegelashvili et al. 2000). Though the initiator is unclear, another pathway has been shown to regulate EAAT2 expression: PI3K phosphorylates Akt, which in turn phosphorylates mTOR, leading to an increase in EAAT2 protein (Wu et al. 2010). A fraction of EAAT2 was found to be constitutively sumoylated in mouse CNS (Foran et al. 2014). The sumoylated transporter is localized to intracellular compartments and promotion of desumoylation led to increased glutamate uptake, indicating a potential target for increasing EAAT2 function. Furthermore, post-transcriptional regulation and external stressors such as oxidative stress can affect EAAT2 expression (Tian et al. 2007; Zagami et al. 2009; Zagami et al. 2005)

Glutamate transporters are decreased and dysregulated in cell models, animal models, and in ALS patients. A 50% decrease in EAAT2 protein was seen in spinal cord homogenates of SOD1-G85R mice (Bruijn et al. 1997). Similar decreases are seen in spinal cord homogenate and the ventral horn of the lumbar spinal cord of SOD1-G93A mice (Bendotti et al. 2001). Focal loss of EAAT2 in the ventral horn of the spinal cord was also seen in the SOD1-G93A rat model of ALS, actually preceding neuronal degeneration. (Howland et al. 2002) Both gain and loss of TBPH (the TDP-43 homolog in Drosophila) altered mRNA expression of EAAT2, suggesting a link between pathological TDP-43 aggregates and mutations and EAAT2 dysregulation (Diaper et al. 2013). TDP-43 also binds to the 3’-UTR of EAAT2 mRNA, but substantially less so in FTLD-TDP brains. (Tollervey et al. 2011) A decrease in EAAT2 expression was also suggested in immunohistochemical studies of rats expressing mutant TDP-43 (M337V) in astrocytes, though the data was not quantified and later control time points were not shown (Tong et al. 2013). Interestingly, FUS has also been found to bind to the 3’ UTR of EAAT2, suggesting regulatory functions (Lagier-Tourenne et al. 2012). While the loss of EAAT2 is believed to contribute to excitotoxicity in ALS, as glutamate cannot be efficiently cleared from the synapse. Upregulation of EAAT2 in the spinal cord of SOD1-G93A mice via intraspinal delivery of an AAV8-Gfa2 vector to the ventral horn at disease onset did not protect phrenic motor neurons, their innervations, nerve function, nor extend lifespan (Li et al. 2015a). This could suggest that the decrease of EAAT2 has its most substantial impact before clinical onset. Similarly, a group II mGlu agonist, LY379268, enhanced EAAT2 expression in the spinal cord and rescued motor neurons, but did not extend the lifespan of SOD1-G93A mice (Battaglia et al. 2015).

Similar, yet more nuanced changes have been observed in human patients as well. In a small patient study, the density of D-aspartate binding, indicative of total glutamate transporter presence and unable to distinguish transporter subtypes, was decreased in the substantia gelatinosa and intermediate gray matter of motor neuron disease patients with a greater decrease in the ALS subgroup (Shaw et al., 1994). EAAT2 specifically was decreased in the motor cortex (71%) and spinal cord (~60%) of ALS patients, with a dramatic decrease (90%) in the motor cortex of a quarter of the patients (Rothstein et al. 1995). Another EAAT2-specific antibody showed that EAAT2 was decreased in the grey matter of the lumbar spinal cord but actually slightly increased in the middle laminae of the motor cortex (Fray et al. 1998). The discrepancy could potentially be explained by differences in the disease progression of patients. EAAT2 was found to decrease in the anterior horns of ALS and LMND patients, correlating with neuronal loss. Sasaki et. al. observed an increase in EAAT2-positive granules in the ventral horn of patients with mild neuronal loss but a decrease in EAAT2 expression in patients with severe neuronal loss (Sasaki et al. 2000).

Recently, purer cultures of iPS-derived astrocytes were generated following insertion of GFP driven by the GFAP promoter using zinc-finger nuclease technology (Zhang et al. 2016). Lines from patients with the ALS-causing mutations SOD1-A4V and C9orf72 demonstrated similar or greater basal expression levels of EAAT2 and glutamate uptake when compared to a control patient line. They also exhibited a more dramatic increase in uptake when co-cultured with neurons. Though only one line of each mutation was studied, the results are suggestive that EAAT2 expression in astrocytes of human patients might not be decreased. EAAT2 expression was also used as a confirmation of astrocyte maturity, so lines expression little or no EAAT2 might not have been included.

In a new cell model for C9orf72-linked ALS, poly-dipeptides of PR, GR, and GA were transfected into NSC-34 cells, to mimic the RAN translation from the poly-hexanucleotide repeat (Kanekura et al. 2016). PR₂₀ induced cell death and inhibited protein translation, as did GR₂₀ to a lesser extent. PR₂₀ interacts with the mRNA of proteins including EAAT2, blocking access of translation factors such as eIF4E and eIF4G, impeding their translation.

It is unlikely that the changes seen in EAAT2 at the protein level are due to DNA mutations or novel RNA variants. The discrepancies could be due to the existence of multiple isoforms of EAAT2, which behave differently in disease yet have not historically been distinguished (Lauriat and McInnes 2007). No gross loss of EAAT2 mRNA was evident in the motor cortex of ALS patients, even those with a dramatic loss of EAAT2 protein (Rothstein et al. 1996). ALS was also not associated with genetic linkage to the EAAT2 gene nor with point mutations within it (Aoki et al. 1998; Jackson et al. 1999).

Dumont et. al. found that the mRNA levels of EAAT2a and EAAT2b isoforms were differentially regulated during disease course (Dumont et al. 2013). While EAAT2a mRNA was 3-fold higher than EAAT2b in the cortex of young SOD1-G93A rats, EAAT2a mRNA decreased by 50% and EAAT2b increased by 50%, until they were at almost equivalent levels in adult and end-stage rats, while no changes were seen in wild-type rats. That change coincided with a decrease in excitatory amino acid uptake by EAAT2 in cortical synaptosomes. In the lumbar spinal cord, EAAT2a increased over time in adult wild-type rats and was reduced in the ventral horns of late-stage mice. EAAT2b was elevated in young SOD1-G93A rats as compared to wild-type controls, but gradually decreased in both the ventral and dorsal horns with disease progression. In impressive agreement, overall EAAT2 levels decreased dramatically in the motor cortex of ALS patients accompanied by a loss in function, yet EAAT2b increased more than two-fold (Maragakis et al. 2004).

Aberrant mRNAs transcripts may play a role, though data have been mixed. Transcripts with partial intron-7 retention and exon 9 deletion were initially found in the motor cortex and spinal cord only of ALS patients and were correlated with EAAT2 protein loss (Lin et al. 1998). Later studies, however, demonstrated the presence of the same aberrant transcripts in the motor cortex of non-neurologically diseased control patients (Meyer et al. 1999). Differential expression of splice variants has been suggested in ALS brain tissue (Honig et al. 2000; Münch et al. 2002). Partial intron 7-retention and exon 9-skipping transcripts led to the production of truncated forms of EAAT2 which were shown in vitro to be rapidly degraded and have a dominant negative effect on normal EAAT2. Similarly, EAAT2 mRNA transcripts were altered over disease course in the SOD1-G93A mouse model (Münch et al. 2002). Later studies, however, found the same variants in normal controls, Alzheimer disease patients and Lewy body dementia patients, albeit at possibly lower (but not statistically significantly so) levels (Honig et al. 2000; Meyer et al. 1999). The same aberrant transcripts were recently produced in a cellular model of the C9ORF72 mutation (Kwon et al. 2014). Human astrocytes expressing 15 or 20 repeats of a synthetic version of the C9ORF72-associated RAN peptide PR expressed the aberrant EAAT2 mRNA forms after only 36 hours, in a repeat-length-dependent manner. Other toxins did not produce the same aberrant splicing. RNA editing of A₁₅₉₁ to I in intron 7 occurs significantly more frequently in the spinal cord, motor cortex, and prefrontal cortex but not cerebellum of sporadic ALS patients than in controls (Flomen and Makoff 2011). Editing activates two cryptic polyadenylation sites, which in turn cause intron-7 retention with termination of transcription transcripts.

Perhaps most indicative, glutamate transport was decreased in synaptosomes derived from the spinal cord (59%), motor cortex (70%), somatosensory cortex (39%) of patients with ALS as compared to healthy and diseased controls (Rothstein et al. 1992). The same decrease was not seen in the unaffected tissues of visual cortex, striatum, and hippocampus, further suggesting a disease-specific decrease. The decrease in transport was likely due to a decrease in overall protein as no change was seen in the affinity of the transporter for glutamate.

Post-translational pathways might also contribute to the loss of EAAT2 protein or function in ALS. Oxidative reactions catalyzed by SOD1-A4V and-I113T but not -WT were found inactivate EAAT2 and inhibit glutamate uptake (Trotti et al. 1999). Caspase-3 activation has been observed in cell and rodent models of ALS while both primary isoforms of EAAT2 contain a caspase-3 consensus sequence (Boston-Howes et al. 2006). Cleavage to a truncated EAAT2 (trEAAT2) and a C-terminal fragment/end (CTE) occurs in a dose- and time-dependent manner and leads to a decrease in glutamate uptake, but is blocked by a D to N mutation in the consensus site. The sumoylated product accumulates in the nucleus of astrocytes with disease progression, though it should be noted that the sumoylation site exists only on the more prevalent EAAT2a isoform. Transfection of an artificially fused CTE-SUMO1 construct induced astrocytes to become toxic to motor neurons via secreted factors (Foran et al. 2011). Expression of SOD1-G93A in MDCK cells led to the internalization of EAAT2 (but not EAAT3) and degradation in acidic compartments and inhibited synthesis. (Vanoni et al. 2004) It is also interesting to note that Riluzole enhances the activity of glutamate transporters including EAAT2 (Fumagalli et al. 2008). While riluzole has multiple effects on the CNS, increased reuptake of glutamate by GluTs could inhibit excitotoxic pathways and contribute to the protective effects seen.

Therapeutic targeting of EAAT2

Due to the role of glutamate dysregulation in ALS and its potential consequence of causing excitotoxicity, already believed to play a substantial role in the progression of ALS, substantial research has been directed towards finding and demonstrating the efficacy of therapeutics that upregulate astrocytic EAAT2 expression. Screening processes have enabled the detection of such compounds, which have shown promise in vitro and in animal models of ALS (Kim et al.2011). Using a library of FDA-approved compounds, researchers have searched for treatments with an already substantial research literature and a faster pathway to clinical trials. Screening strategies have employed immunoblotting of spinal cord slice cultures, ELISA of astrocyte-like PA-EAAT2 cells, and firefly lucieferace expression on human fetal derived immortalized astroglia using the E2 promoter (Colton et al. 2010; Li et al. 2015b; Li et al. 2011).

In one of the first such screens for an EAAT2 upregulator, organotypic rat spinal cord slices were treated with a library of 1040 FDA-approved compounds (Rothstein et al. 2005). Homogenates of the treated tissues were blotted for EAAT2 protein expression, yielding numerous hits. A disproportionate number of the top 2% of hits fell in the class of β-lactam antibiotics, which were found to increase protein via enhanced promotor activity. Already known to penetrate the blood brain barrier and to not cause substantial CNS toxicity, ceftriaxone (a β-lactam antibiotic) was found to increase EAAT2 protein expression in treated mice, both substantially and persistently. SOD1-G93A mice treated before onset had a decreased motor neuron loss after two weeks. When ceftriaxone was administered daily to SOD1-G93A mice beginning approximately at the time of onset, a delay in the loss of muscle strength and body weight was seen for 4–6 weeks and a 10-day extension of survival was found. Further suggesting promise as a clinical therapeutic, ceftriaxone increases NF-κB binding to the EAAT2 promoter in primary human fetal astrocytes, increasing transcription (Lee et al. 2008).

Despite significant promise from the aforementioned preclinical studies and early phase clinical trials, a phase 3 clinical trial showed that ceftriaxone was not effective for treating ALS. In the phase I branch of the study, 66 patients were randomized to placebo (pediatric multivitamin), 2 g/day, or 4 g/day via central venous catheter (Berry et al. 2013). Ceftriaxone was shown to remain above the target 1 uM in the CSF with both dosages, indicating sufficient penetration. The same patients were followed for 20 weeks, demonstrating sufficient safety and tolerability, other than hepatobiliary adverse events treated with ursodeoxycholic acid (which was given to all ceftriaxone patients in phase 3). Promisingly, functional decline was slower in patients in the high dose group versus the placebo (p=0.0416) (Cudkowicz et al. 2014). Unfortunately, the decrease in functional decline was not replicated in the stage 3 interim or final analysis, nor was a difference seen in survival. Furthermore, adverse event rates were significantly and substantially higher in the treatment group for gastrointestinal, hepatobiliary, and blood or bone marrow related events. Participants receiving ceftriaxone also had more hepatobiliary serious adverse events but fewer infection-related serious adverse events.

Though ceftriaxone was ineffective in treating ALS patients, it remains unclear whether EAAT2 upregulation is no longer a viable clinical target: The actual efficacy of ceftriaxone in increasing EAAT2 in the CNS was not determined as no markers existed at the time of the study, though a PET ligand for EAAT2 is being developed (Gerdes et al. 2015). Animal models also demonstrated efficacy of ceftriaxone only when administered before symptom onset. Perhaps ceftriaxone would be more valuable for pre-onset familial ALS carriers. Stratification by mutation status, onset type, or cognition status might also yield interesting results (van den Berg 2014).

In separate studies, beginning with a cell-based ELISA screen of 140,000 small molecules compounds, Colton et al identified 293 compounds that increased the expression of EAAT2 in PA-EAAT2 cells by at least 70% (3 standard deviations), 61 of which demonstrated a dose-dependent effect. Of those, 3 were selected for further optimization due to their potency, lack of toxicity, and following confirmatory biochemical and functional studies (Colton et al. 2010). A representative compound of a pyridazine-based series derived from those experiments, LDN/OSU-0212320 was further characterized in a mouse model (Kong et al. 2014). The compound protected motor neurons co-cultured with astrocytes from glutamate-induced excitotoxicity. Furthermore, with treatment beginning near onset, the compound slowed motor decline and weight loss and substantially extended survival, suggesting that it is a promising agent for future trials.

Glutamate uptake by a clonal neural hybrid cell line (MN-1) known to express GluTs was assessed after treatment by a library of 1040 FDA-approved compounds (Boston-Howes et al. 2008). NDGA was identified by the screen and found to enhance glutamate transport in a dose-dependent matter and increase EAAT2 transport by 3-fold. While it also increased glutamate uptake in vivo as measured in synaptosomes of treated mice, it failed to increase glutamate uptake in symptomatic SOD1-G93A mice or to extend their lifespan.

Using an immortalized astrocytic cell line expressing firefly luciferace on the EAAT2 promotor, Li et. al. screened a library of 1040 FDA-approved compounds for EAAT2 upregulation (Li et al. 2011). Harmine, a naturally occurring beta-carboline alkaloid and one of the top hits, was subsequently found to induce EAAT2 and EAAT1 mRNA and protein expression (as well as expression of their counterparts in mouse-derived cells). Furthermore, treatment of early onset SOD1-G93A mice led to an increase of EAAT2 in the cortex.

Other recent preclinical trials have yielded promising results in the treatment of ALS by EAAT2 modification. Benkler et. al. treated SOD1-G93A mice with a cocktail of lentiviruses encoding EAAT2, GDH2, and NRF2, with the objectives of increasing EAAT2 expression, reducing glutamate availability, and minimizing oxidative stress, respectively. While all three genes led to some neuroprotection in vitro, the combination led to the most neuroprotection and it was found to expand lifespan by 19–22 days after administration at 65 days of age (Benkler et al. 2015). This finding further suggests that EAAT2-modifying therapies might be more effective in combination with other treatments.

An alternative explanation for the failure of ceftriaxone in clinical trials and of EAAT2-upregulating drugs in symptomatic animal models is the existence of a secondary EAAT2-mediated neurotoxicity pathway: As previously mentioned, EAAT2 is cleaved by caspase-3 in ALS, leading to the generation of CTE, which was found to be sumoylated. (Gibb et al. 2007) Expression of an artificially fused CTE-SUMO1 causes astrocytes to secrete factors toxic to motor neurons (Foran et al. 2011). Not only does that suggest an additional pathway for EAAT2 toxicity in ALS, but that pathway would theoretically be upregulated with many of the EAAT2-targeted treatments, potentially preventing their success by coupling a decrease in excitotoxicity with an increase in secreted toxic factors.

Conclusions

Despite decades of research and advances in our understanding of ALS, riluzole remains the only FDA-approved disease altering medication and prolongs life by only a few months. While it has become apparent that many converging pathways are at work, excitotoxicity remains a probable contributor to motor neuron death. The dysregulation of the glutamate transporter EAAT2 presumably leads to increased synaptic glutamate, causing excessive glutamate signaling and death of post-synaptic neurons. EAAT2 is also cleaved by caspase-3 in ALS, leading the nuclear accumulation of a sumoylated fragment and astrocytic secretion of toxic factors.

Substantial research efforts have been directed towards transcriptional, translational, and functional activators of EAAT2 as therapeutics for ALS. While they have shown great promise in vitro and in pre-clinical animal models, they have yet to demonstrate efficacy in human patients. New tools which should aid in the testing of EAAT2-modifying agents are in development. While ceftriaxone failed in clinical trials, it remains unknown whether or not it achieved is mechanistic goal of upregulating EAAT2. A novel PET tracer for EAAT2 is being developed which would enable detection of EAAT2 changes in human patients throughout treatment (Gerdes et al. 2015). Such an agent would allow researchers to determine whether ceftriaxone, or other therapeutic agents, actually upregulate EAAT2 before asking if EAAT2 upregulation extends lifespan in human patients.

Mutations in SOD1 account for only 20% of familial ALS cases and 5% of sporadic cases, while most therapeutics are tested in SOD1 mutant models (Robberecht and Philips 2013). The creation of novel models based on the substantially more prevalent C9ORF72 repeat expansion and on iPS models derived from sporadic patients may enable more efficient and predictive screening of ALS therapeutics, including EAAT2-upregulating drugs (van Blitterswijk et al. 2012; Dimos et al. 2008). Motor neurons, astrocytes, and fibroblasts derived from iPSCs of ALS patients are becoming an increasingly reliable model of disease and degeneration. Libraries of cells derived from familial and sporadic patients are being developed (Li et al. 2015b). These should enable more humanized models of disease, such as examination of EAAT2 expression and function and excitotoxicity across the spectrum of sporadic ALS. Unfortunately, EAAT2 is frequently used as a marker for maturation of astrocytes from iPSCs, which could confound studies of its expression and functional levels (Li et al. 2015b; Zhang et al. 2016). Despite the setback of ceftriaxone’s failure as a therapeutic, EAAT2 remains a viable and promising target for the treatment of ALS.